Lin-Han Jia @ LAMDA, NJU-CS
|
贾林瀚 Lin-Han Jia
Ph.D. Student Supervisor: Professor Yu-Feng Li Email: jialh@lamda.nju.edu.cn Tel/WeChat: 15513127317 Laboratory: Room 912, Computer Science Building, Xianlin Campus of Nanjing University [GitHub][Google Scholar][ZhiHu] |
|
Biography
2022-present, Ph.D. student in the School of Computer Science at Nanjing University, advised by Prof. Yu-Feng Li, joining the Machine Learning and Data Mining (LAMDA) Group founded by Academician Zhi-Hua Zhou.
2018-2022, B.Sc. in the College of Computer Science and Technology at Jilin University, joining the Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education founded by Academician Xianghao Wang.
Research Interests
My research primarily focuses on self-supervised, semi-supervised, weakly-supervised, and unsupervised learning, while delving deeply into neuro-symbolic, reasoning, abductive, and verification in knowledge-data dual-driven learning paradigms, with emphasis on fundamental research for leveraging unlabeled data and prior knowledge in label-limited scenarios.
Specific interests include:
- Intrinsic correlations between pre-training and downstream tasks: Clarifying capability boundaries of various unsupervised pre-training methods, such as vision models trained via contrastive learning and masked reconstruction, language models trained via autoregressive and autoencoding (diffusion) approaches, and multimodal models trained via modality matching and generation, in relation to their limitations in downstream objectives and connections to pre-training tasks themselves.
- Quantifiable training decision framework: Establishing a quantitative framework for model training decision processes to estimate training decision benefits before high-cost training, reducing time and computational costs of large-scale training.
- Formal modeling of learning objectives: Precisely guiding model training through formal modeling of downstream demand distributions, reducing ambiguity and trial-and-error in model training objectives.
- Training pipeline reasoning: Promoting the transition of training framework automated design from search paradigms to reasoning paradigms, achieving downstream task-oriented adaptive combination design of training tasks.
- Stability of controllable model semantic distributions: Ensuring stability of model output semantic distributions under prior knowledge constraints, balancing compliance constraints with preservation of original semantics.
- Substitutability of prior knowledge for labels: Exploring the relationship between knowledge base completeness and label substitution, clarifying core conditions for knowledge systems to replace massive labels.
- Construction of self-evolving prior knowledge bases: Attributing prior knowledge for unknown tasks, building knowledge bases adaptable to full model training processes with autonomous iterative capabilities.
- Joint optimization of knowledge and neural systems: Overcoming consistency challenges in fusing discrete knowledge systems with continuous neural networks, optimizing combinatorial optimization algorithms for exponential solution spaces and reducing computational complexity, while exploring the potential of quantum algorithms in combinatorial optimization problems.
Publications
Conference Papers
Lin-Han Jia, Si-Yu Han, Wen-Chao Hu, Jie-Jing Shao, Wen-Da Wei, Zhi Zhou, Lan-Zhe Guo, Yu-Feng Li. Quantitative Estimation of Target Task Performance from Unsupervised Pretext Task in Semi/Self-Supervised Learning. In: Proceedings of the 43-rd International Conference on Machine Learning (ICML'26). 2026.[Paper][Code]
Hong-Jie You, Jie-Jing Shao, Xiao-Wen Yang, Lin-Han Jia, Lan-Zhe Guo, Yu-Feng Li. Pianist Transformer: Towards Expressive Piano Performance Rendering via Scalable Self-Supervised Pre-Training. In: Proceedings of the 43-rd International Conference on Machine Learning (ICML'26). 2026.[Paper]
Wen-Chao Hu, Qi-Jie Li, Lin-Han Jia, Cunjing Ge, Yu-Feng Li, Yuan Jiang, Zhi-Hua Zhou. Curriculum Abductive Learning. In: the 39-th Annual Conference on Neural Information Processing Systems (NeurIPS' 25). 2025.[Paper]
Shi-Yu Tian, Zhi Zhou, Kun-Yang Yu, Ming Yang, Lin-Han Jia, Lan-Zhe Guo, Yu-Feng Li. VC Search: Bridging the Gap Between Well-Defined and Ill-Defined Problems in Mathematical Reasoning. In: Proceedings of the 30-th Conference on Empirical Methods in Natural Language (EMNLP'25). Oral Presentation. 2025.[Paper]
Si-Yu Han, Lin-Han Jia, Yu-Feng Li. 基于正则的稳健数据重载量子学习模型. In: Proceedings of the 20-th Chinese Conference on Machine Learning (CCML'25). Best Paper Award Nomination. 2025.
Lin-Han Jia, Wen-Chao Hu, Jie-Jing Shao, Lan-Zhe Guo, Yu-Feng Li. Verification Learning: Make Unsupervised Neuro-Symbolic System Feasible. In: Proceedings of the 42-nd International Conference on Machine Learning (ICML'25). 2025.[Paper][Code]
Xiao-Wen Yang, Jie-Jing Shao, Lan-Zhe Guo, Bo-Wen Zhang, Zhi Zhou, Lin-Han Jia, Wang-Zhou Dai, Yu-Feng Li. Neuro-Symbolic Artificial Intelligence: Towards Improving the Reasoning Abilities of Large Language Models. In: Proceedings of the 34-th International Joint Conference on Artificial Intelligence (IJCAI '2025 Survey Track). 2025.[Paper]
Lin-Han Jia, Lan-Zhe Guo, Zhi Zhou, and Yu-Feng Li. Realistic Evaluation of Semi-Supervised Learning Algorithms in Open Environments. In: Proceedings of the 12-th International Conference on Learning Representations (ICLR'24). Spotlight Presentation. 2024.[Paper][Code]
Zhi-Zhou, Lan-Zhe Guo, Lin-Han Jia, Ding-Chu Zhang, Yu-Feng Li. ODS: Test-Time Adaptation in the Presence of Open-World Data Shift. In: Proceedings of the 40-th International Conference on Machine Learning (ICML'23). Oral Presentation. 2023.
Lin-Han Jia, Lan-Zhe Guo, Zhi Zhou, Jie-Jing Shao, Yuke Xiang, Yu-Feng Li. Bidirectional Adaptation for Robust Semi-Supervised Learning with Inconsistent Data Distributions. In: Proceedings of the 40-th International Conference on Machine Learning (ICML'23). Oral Presentation. 2023.[Paper][Code]
Journal Papers
Lan-Zhe Guo, Lin-Han Jia, Jie-Jing Shao, Yu-Feng Li. Robust Semi-Supervised Learning in Open Environments. Frontiers of Computer Science. 2025.[Paper]
Lin-Han Jia, Lan-Zhe Guo, Zhi Zhou, and Yu-Feng Li. LAMDA-SSL: A Comprehensive Semi-Supervised Learning Toolkit. Science CHINA Information Science. 2023. [Paper][Code][Slide]
Preprint
Lin-Han Jia, Jie-Jing Shao, Wen-Chao Hu, Zhi Zhou, Lan-Zhe Guo, Yu-Feng Li. Learning Task Design Based on Fuzzy Logical Reasoning. 2026.
Wen-Chao Hu, Lin-Han Jia, Cunjing Ge, Yu-Feng Li, Yuan Jiang, and Zhi-Hua Zhou. ABL-Forest: Localized Abductive Learning with Formal Guarantees. 2026.
Rui-Zhi Qin, Jie-Jing Shao, Xiao-Wen Yang, Lin-Han Jia, Lan-Zhe Guo, Yu-Feng Li. COAST: Self-Correction Learning for Mastering Constraint Satisfaction in Travel Planning. 2026.
Hong-Jie You, Jie-Jing Shao, Xiao-Wen Yang, Zhi-Fan Wu, Lin-Han Jia, Lan-Zhe Guo, Yu-Feng Li. FormalImG: Evaluating Structural Compositional Generalization for T2I Models. 2026.
Lin-Han Jia, Si-Yu Han, Lan-Zhe Guo, Zhi Zhou, Zhao-Long Li, Yu-Feng Li, Zhi-Hua Zhou. A Smooth Transition Between Induction and Deduction: Fast Abductive Learning Based on Probabilistic Symbol Perception. arXiv preprint arXiv:2502.12919. 2025.[Paper]
Lin-Han Jia, Lan-Zhe Guo, Zhi Zhou, Si-Yu Han, Zi-Wen Li, Yu-Feng Li. Detecting Scarce and Sparse Anomalous: Solving Dual Imbalance in Multi-Instance Learning. arXiv preprint arXiv:2503.13562. 2025.[Paper]
Si-Yu Han, Lin-Han Jia, Lan-Zhe Guo. Multiple Embeddings for Quantum Machine Learning. arXiv preprint arXiv:2503.22758. 2025.[Paper]
Software
|
Awards & Honors
CCML Best Paper Award Nomination, 2025
LAMDA Elite Award, 2025
Outstanding Graduate Student, 2025, 2024 and 2023
Elite PhD Scholarship, 2024
National Scholarship, 2023
Outstanding Graduate, 2022
Outstanding Undergraduate Thesis, 2022
CCF Elite Collegiate Student Award, 2021
Outstanding Student, 2021 and 2020
ACM International Collegiate Programming Contest (ICPC) Asia Regional Contest, Silver Medal, 2021
China Collegiate Programming Contest (CCPC) Provincial Contest, Gold Medal, 2020
Mathematical Contest In Modeling (MCM), Finalist Prize, 2020
China Undergraduate Mathematical Contest in Modeling (CUMCM) Provincial Contest, First Prize, 2020 and 2019
Academic Services
- Senior Program Committee Member: IJCAI 2025
- Program Committee Member: NeurIPS 2023–2025, AAAI 2024–2026, ICLR 2024–2026, IJCAI 2024–2026, ICML 2025–2026, KDD 2024–2026, CVPR 2026, ECCV 2026
- Conference Volunteer: ICLR 2024, MLA 2022–2023
- Journal Reviewer: Machine Learning, Neurocomputing, Neural Networks
- Teaching Assistant: Advanced Machine Learning (Graduate Course, Nanjing University), 2024–2025
Correspondence
Email: jialh@lamda.nju.edu.cn
Tel/WeChat: 15513127317
Address: Lin-Han Jia, National Key Laboratory for Novel Software Technology, Nanjing University, Xianlin Campus, 163 Xianlin Avenue, Qixia District, Nanjing 210023, China (南京市栖霞区仙林大道163号, 南京大学仙林校区, 软件新技术国家重点实验室, 210023.)